基于灰色阈值模型风险价值的贝叶斯估计.pdfVIP

DevelopmentofEnergyScience November2014，Volume2．Issue4，PP．24·33 BayesianEstimationofValue--at--riskBasedon GrayPeaksoverThreshold Ru幻ingWang DepartmentofSoftwareEngineering，HainanCollegeofSoftwareTechnology，QionghaiHainan571400，China Email：ayrqwang@163．corn Abstract A two-stagemodelforestimatingvalue-at-riskbasedongreysystem andextremevaluetheoryisproposed．Firstly,inorderto capturethedependencies，seasonalitiesnadvolatility-clustering，allGM(1，2)modelisusedtofilterelecrticitypriceseries．Inthis way,anapproximatelyindependentlynadidenticallydisrtibutedresidualserieswithbeaerstatisticalpropertiesisacquired．Then peaksoverthresholdisadoptedtoexplicitlymodeltheatilsoftheresidualsofGM(1，2)model，andaccurateestimatesof elecrticitymraketvalue—at-riskcanbeproduced．Forconqueringthedifficulyt lackingofrsampledaat overhtreshold，Bayesina estimationbasedonMrakovChainM onteCarlosimulationisusedtoestimatethepraametersofpeaksoverthresholdmode1．Th e empiricalnaalysisshowshtattheproposedmodelCna berapidlyreflecthtemostrecentnadrelevantchnagesofelecrticiyt prices andcanproduceaccurateforecastsofvalue-at-riskat allconfidencelevels，nadhtecomputationalcostisfarlesshtna theexisting two-stagevalue-at-riskestimatingmodels，furtherimprovinghteabilityofriskmnaagementofrelectriciytmarketparticipnats． Keywords：Value-at-risk；GreySystemTheory；ExtremeValueTheory；GM (1。2)；PeaksOverThresholds；BayesianEstimation 基于灰色阈值模型风险价值的贝叶斯估计术 王瑞庆 海南软件职业技术学院软件工程系，海南 琼海 571400 摘 要 ：基于现货电价具有信息不完全和不确定的特征，提出了一个基于灰色系统和极值理论的两阶段风险价值计算模 型。该模型首先采用灰色 GM(1，2]模型对电价序列进行过滤，以获得统计特性更好的独立同分布残差序列，然后运用极 值理论的阈值模型直接拟合残差序列的尾部分布，从而获得准确有效的风险价值估计结果。采用基于马尔可夫链蒙特卡 罗模拟的贝叶斯方法估计阈值模型的参数，克服 了超阈值样本数据匮乏的问题。实证分析表明：该模型能对现货电价的 变化做 出迅速的反应，风险价值的估计结果在各置信水平下均准确有效，其计算工作量远小于现有的两阶段风险价值计 算模型